COMPARATIVE USE OF ARTIFICIAL NEURAL NETWORKS FOR THE DISCRIMINATION OF THE WATER RESERVOIRS OF ATHENS ACCORDING TO THEIR ELEMENTAL CONTENT